Mapping the gene space at single-cell resolution with gene signal pattern analysis.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Existing single-cell sequencing analysis methods focus on mapping cellular states, with limited development in gene space embeddings.
• Understanding gene relationships and patterns is crucial for deciphering complex biological systems.

Purpose of the Study:

• To formulate the gene embedding problem and develop a novel computational approach for learning gene representations.
• To evaluate the performance of gene embeddings on simulated single-cell data and establish baseline comparisons.
• To demonstrate the utility of gene representations for diverse biological applications.

Main Methods:

• Formulation of the gene embedding problem and design of evaluation tasks using simulated single-cell data.
• Development of gene signal pattern analysis (GSPA), a graph signal processing approach utilizing diffusion wavelets on a cell-cell graph.
• Establishment of ten baseline methods for comparison.

Main Results:

• GSPA learns rich gene representations by analyzing gene patterning and localization on the cellular manifold.
• Demonstrated efficacy of GSPA for capturing gene co-expression modules, condition-specific enrichment, and perturbation-specific gene-gene interactions.
• Showcased broad utility of GSPA-derived gene representations in cell-cell communication, spatial transcriptomics, and patient response analyses.

Conclusions:

• Gene signal pattern analysis (GSPA) provides a powerful framework for learning gene embeddings from single-cell data.
• GSPA enhances the characterization of gene behavior and relationships within the cellular context.
• The developed gene representations have wide-ranging applications in advancing biological research and clinical insights.